amd64/darwin.
The stack appears to be the m0 OS stack, since it is not allocated and the fmt test does
not use cgo (so it's not some other thread's stack).  Perhaps we are setting the limits
wrong for the m0 stack.  Or perhaps it is an unexpected signal.  Regardless, I'd like to
understand this.



---------- Forwarded message ----------
From: Jonathan Pittman <jonathan.mark.pittman@gmail.com>
Date: Fri, Apr 27, 2012 at 07:54
Subject: Re: [go-nuts] Re: go1.0.1 released
To: Peter Kleiweg <pkleiweg@xs4all.nl>
Cc: golang-nuts <golang-nuts@googlegroups.com>


Not sure if anyone else saw this.  With respect to the new go1.0.1 release, I just
ran...

hg pull
hg update release
./all.bash

and then in the middle of the "# Testing packages." section, I saw...


runtime: split stack overflow: 0x2747cd8 < 0x2748000
throw: runtime: split stack overflow

goroutine 1 [chan receive]:
main.(*builder).do(0xf8400fe180, 0xf8401d5000, 0xf8401bb000, 0x1)
/Users/jonathanpittman/go/src/cmd/go/build.go:606 +0x3e4
main.runTest(0x45e598, 0xf840001480, 0x300000003, 0xf8400f6880)
/Users/jonathanpittman/go/src/cmd/go/test.go:367 +0x1aa1
main.main()
/Users/jonathanpittman/go/src/cmd/go/main.go:141 +0x426

goroutine 2 [syscall]:
created by runtime.main
/Users/jonathanpittman/go/src/pkg/runtime/proc.c:221

<some go routines omitted for brevity>

goroutine 342 [syscall]:
syscall.Syscall()
/Users/jonathanpittman/go/src/pkg/syscall/asm_darwin_amd64.s:14 +0x5
syscall.Read(0x6, 0xf840138200, 0x20000000200, 0x0, 0x0, ...)
/Users/jonathanpittman/go/src/pkg/syscall/zsyscall_darwin_amd64.go:810 +0x78
os.(*File).read(0xf8401d4b78, 0xf840138200, 0x20000000200, 0x1c1460, 0x0, ...)
/Users/jonathanpittman/go/src/pkg/os/file_unix.go:163 +0x58
os.(*File).Read(0xf8401d4b78, 0xf840138200, 0x20000000200, 0x0, 0x0, ...)
/Users/jonathanpittman/go/src/pkg/os/file.go:95 +0x83
bytes.(*Buffer).ReadFrom(0xf84080fae0, 0xf840104600, 0xf8401d4b78, 0x0, 0x0, ...)
/Users/jonathanpittman/go/src/pkg/bytes/buffer.go:153 +0x199
io.Copy(0xf84013dba0, 0xf84080fae0, 0xf840104600, 0xf8401d4b78, 0x0, ...)
/Users/jonathanpittman/go/src/pkg/io/io.go:340 +0xc2
os/exec._func_003(0xf84019b050, 0xf8401d4b70, 0x78133, 0x1c1888, 0x445ab, ...)
/Users/jonathanpittman/go/src/pkg/os/exec/exec.go:201 +0x5a
os/exec._func_004(0xf8401d4b50, 0xf840264900, 0x0, 0x0)
/Users/jonathanpittman/go/src/pkg/os/exec/exec.go:260 +0x1e
created by os/exec.(*Cmd).Start
/Users/jonathanpittman/go/src/pkg/os/exec/exec.go:261 +0x65e
go build fmt: exit status 2

and a few lines further down it had...

FAIL fmt [build failed]

I ran all.bash again, but could not reproduce the problem.  Everything built and tested
fine.  Thought I would mention it in case it comes up again or is related to some
obscure bug.

On Fri, Apr 27, 2012 at 5:19 AM, Peter Kleiweg <pkleiweg@xs4all.nl> wrote:
>
> Will there be a release 'weekly' with these changes? I am using some
> of the experimental packages, but would also like to have the updates
> from go1.0.1.